Ultra-compact on-chip LED collimation opticsby 3D femtosecond direct laser writingSIMON THIELE,1,* TIMO GISSIBL,2 HARALD GIESSEN,2 AND ALOIS M. HERKOMMER1

1Institut für Technische Optik and Research Center SCoPE, University of Stuttgart, Stuttgart, Germany24th Physics Institute and Research Center SCoPE, University of Stuttgart, Stuttgart, Germany*Corresponding author: thiele@ito.uni‑stuttgart.de

Received 18 April 2016; revised 31 May 2016; accepted 1 June 2016; posted 1 June 2016 (Doc. ID 263376); published 27 June 2016

By using two-photon lithographic 3D printing, we demon-strate additive manufacturing of a dielectric concentratordirectly on a LED chip. With a size of below 200 μm indiameter and length, light output is increased by a factorof 6.2 in collimation direction, while the emission half-angleis reduced by 50%. We measure excellent form fidelityand irradiance patterns close to simulation. Additionally, amore complex shape design is presented, which exhibits anonconventional triangular illumination pattern. The intro-duced method features exceptional design freedoms whichcan be used to tailor high-quality miniature illuminationoptics for specific lighting tasks, for example, endos-copy. © 2016 Optical Society of America

OCIS codes: (220.2945) Illumination design; (220.4000)

Microstructure fabrication; (220.3740) Lithography; (230.3670)

Light-emitting diodes.

http://dx.doi.org/10.1364/OL.41.003029

Additive manufacturing of optical components in threedimensions is a fast-growing research field [1–8]. Differentapproaches, such as inkjet printing or stereolithography areused, and corresponding printers are commercially available.With resolutions of <100 nm and optically clear materials, di-rect laser writing by two-photon lithography is ideally suited forthe fabrication of optical elements in the micrometer regime [2].A variety of different micro- and nano-optical components, cre-ated by femtosecond direct laser writing, have been successfullydemonstrated in recent years. These include photonic crystals[3], optical waveguides [4], diffractive elements [5], and refrac-tive freeform surfaces [6], as well as hybrid diffractive-refractiveelements [7], and even multi-component optical systems [8,9].

Due to the simple processing requirements, femtoseconddirect laser writing is compatible with a variety of substrates.The chemical structure of the resists enables strong adhesionon many different surfaces. Optical elements have been manu-factured directly on optical fibers [10] or integrated photoniccircuit chips [4]. The direct fabrication on optoelectronic sur-faces such as emitting areas of light emitting diodes (LED) orvertical cavity lasers (VCSELS), as well as on the pixel matrix ofimage sensors, is easily feasible.

LED chips today are crucial components for many opticalapplications and can be miniaturized down to the micrometerscale. However, influencing the light output distribution onthe same scale has proved challenging. Integrated photonic struc-tures can change the radiation characteristics of bare LED chipsand decrease their étendue [11]. Similarly, chip-integrated reflec-tive sidewalls are used to decrease the angular extent of the source[12], although only by small amounts. Refractive primary opticsthat transform the mostly Lambertian emission patterns usuallyconsist of simple domes with diameters considerably greater thanthe active area of the chip. Schreiber et al. reported on diamondturned concentrators as primary optics [13], although with di-mensions roughly a factor of ten above the optics presented here.An example of collimation optics by femtosecond direct laserwriting was published by Jiang et al. [14]. However, in this case,a laser diode with high spatial coherence was used. This requires adifferent design approach for collimation compared to our LEDcase and does not require an undercut design, as is presented here.

Efficient primary optics play a crucial role for the miniaturi-zation of LED illumination systems. In principle, two types ofprimary optics can be distinguished. In the first case, the lens isplaced with an air gap above the chip. This helps in étendue-critical applications because étendue is not raised due to ahigher index material being in contact with the chip [15].In the second case, a material with a refractive index higherthan air is in contact with the LED chip.

This increases the overall efficiency because less total inter-nal reflection (TIR) happens at the boundary between the chipwith its high refractive index and the outside world.

A miniaturized primary optics can significantly reduce the sizeand shape requirements for the secondary optics and, thus, help todecrease the total system dimensions. In this Letter, we demon-strate the usefulness of additive manufacturing to create a newclass of on-chip LED optics with unprecedented design freedomand unmatched optical properties on this length scale. In the fol-lowing, the design, fabrication, and experimental characterizationare described.

The ray tracing design starts with modeling of the LEDsource. For the experiments, an InGaAs point source chip(OSRAM F1372B) with a circular emitting area of 80 μmin diameter was used. The center wavelength is 650 nm,and the emission pattern is Lambertian. The star-shaped Au

Letter Vol. 41, No. 13 / July 1 2016 / Optics Letters 3029

0146-9592/16/133029-04 Journal © 2016 Optical Society of America

electrodes [Fig. 2(a)] are considered in the model with a reflec-tivity of 100%. With our fabrication approach, the diameter ofthe optical element cannot exceed 200 μm, which limits theachievable collimation. In the case of conservation of energy,the light output angle is given by

β � sin−1�n1d 1 sin�α�

n2d 2

�; (1)

in dependence of the inside and outside refractive indices n1and n2, input diameter d 1, output diameter d 2, and the inputangle α (Fig. 1). In this Letter, the refractive index n1 of theused photoresist was measured to be 1.5101 at 650 nm.Considering a half-angle of 60° at FWHM radiant intensityfor the Lambertian emitter, the best achievable half-angle ofthe current setup in air is thus 31.4°, if étendue is conserved.

Because of the broad angular spectrum of the LED chip, adielectric total internal reflecting concentrator (DTIRC) [16]was chosen as a first stage of collimation. An aspheric lens

on top provides additional degrees of freedom to influencethe final light output. While the DTIRC was designed fromanalytical considerations, the aspheric top surface was itera-tively optimized to achieve a maximum optical power at a pre-defined target area and to correct for effects of the electrodepattern. The shape of the DTIRC part is defined accordingto [17] through the coordinate r in radial direction as afunction of coordinate z along the symmetry axis [Eq. (2)]:

r � −b�ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffib2 − 4ac

p

2a; where

a � C2;

b � 2�CSz � r0P2�;c � z2 S2 − 2r0CQz − r20PT ;

C � cos�θ�; S � sin�θ�; P � 1� S;

Q � 1� P; and T � 1� Q: (2)

The two parameters, emission radial aperture r1 and maxi-mum top angle β, are set to 39 μm and 15°, respectively.

The top aspheric surface is defined as a function ofradius r as

z � cr2

1�ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1 − c2r2

p � a2r2 � a4r4 � a6r6; (3)

with a curvature c of −2.858 · 10−3 μm−1 and the asphericcoefficients a2 � −9.88 · 10−4 μm−1, a4 � −2.26 · 10−7 μm−3,and a6 � −1.46 · 10−13 μm−5. Figure 1 depicts ray tracing plotsfor the source, as well as for the combined system.

To facilitate testing, the LED chips are processed on TO-18metal packages. The femtosecond direct laser writing is carriedout using the commercially available lithography systemPhotonic Professional GT from the company NanoscribeGmbH. The photoresist IP-S, specifically developed forsmooth surfaces, is used [18].

Details about the dip-in lithography process can be foundin [9]. Figure 2 shows the chip before and after processing. InFig. 3 the final structure is compared to a standard LED in atraditional package with a base diameter of 5 mm.

Fig. 1. (a) Illustration of Eq. (1). (b) Cross-sectional view of randomsource rays, emitted from the bare chip. (c) Effects of the designedcompound parabolic concentrator collimator.

Fig. 2. Fabrication results. (a) LED chip before processing. (b) LEDafter processing. (c), (d) Chromatic-confocal surface measurement ofthe top lens surface.

Fig. 3. Size comparison of the 3D printed compound parabolic con-centrator lens on top of a chip in a TO-18 package with a conventional5 mm LED in traditional package.

3030 Vol. 41, No. 13 / July 1 2016 / Optics Letters Letter

As visible in the microscope images, the resulting structureexhibits a smooth surface appearance similar to glass. The shapefidelity after fabrication is characterized by measurement ofthe top aspheric surface with a chromatic-confocal sensor.Comparison between design and measurement exhibits goodagreement. AFM measurements from previous work reveal aRMS roughness of ∼13 nm on comparable lenses, fabricatedwith the same method and similar parameters. On steeperflanks, the staircase effect of the layer-by-layer writing is stillvisible, an effect which can be reduced by smaller layer spacingor a contour-based writing algorithm.

In this case, the layer spacing is set to 100 nm as trade-offbetween writing time and accuracy. The total time to write thestructure is about 3.5 h. There are a number of possibilities toincrease the writing speed. For example, the process time can bereduced by writing a shell only and flood exposing the sampleafterward [7].

Before and after fabrication, the LEDs are tested with aforward current of 10 mA, and the irradiance distribution isrecorded. For the measurement, the LED is placed below a

diffuser plate with Lambertian scattering properties [Fig. 4(a)].The distance between LED and diffuser is chosen far enoughto ensure that the emission area of the optics (diameter:200 μm) can be considered as point source.

The light distribution on the diffuser is imaged with astandard DSLR camera, while the lateral scale is determinedby imaging a ruler.

The raw data images are recorded and converted to a dataarray after linear pixel readout and merging the four RGGB pix-els. For comparison with simulation, the irradiance distributionon a detector at the same distance as in the experiment is calcu-lated by ray tracing. Effects such as multiple Fresnel reflectionsand reflections at the Au electrodes are taken into account.

The results of simulation and experiment are shown inFigs. 4(b)–4(e). If light output is compared in forward direc-tion, the DTIRC structure shows an increase in irradiance by afactor of 6.2 compared to the bare chip. This value is consid-erably higher than initially proposed by simulation (3.01). Thereason for this difference lies in the fact that a higher indexmaterial on the emission area of the chip leads to an overallincrease in light extraction efficiency compared to air.

The increase can be calculated according to Eq. (4), whichconsiders the change in critical TIR angle inside the chip independence of the two refractive indices nLED and nResist [19]:

ηIP-Sηair

�1 − cos

�asin

�nResistnLED

��

1 − cos�asin

�1

nLED

�� : (4)

While the refractive index of the polymerized photoresist nIP-Sis known to be 1.5101, it was not possible to determine therefractive index of the LED top layer due to confidentiality.It is reasonable however, to assume GaAs as material [20],which has a refractive index of 3.83 at λ � 650 nm. Thiswould lead to an increase in extracted power by a factor of2.33. Combined with the increase by collimation, a factorof 6.6 results, which is not far from the measured value of6.2. The remaining difference can be explained by scatteringlosses at writing layers and other imperfections of the structure.

In general, simulation and experiment match well. For bet-ter comparison, polar plots of radiant intensity as a function ofangle are displayed in Fig. 4(e). The values were calculated bymultiplying the irradiance with the squared Cartesian distancebetween source and point of interest and dividing the results bythe cosine of the polar angle [21]. While the simulation and theexperiment show similar Lambertian distributions in the case ofthe bare chip, the curves differ more visibly in the case of thecollimation optics. This effect can be explained by deviationsfrom design in the DTIRC shape. The measured FWHM half-angles [Fig. 4(e)] of radiant intensity are very close to the theo-retical limit of 31.4° in the case of conservation of energy. Theobservation that the value seems to be slightly below this limitindicates that the designed efficiency is not fully reached.

The high degree of design freedom of the presented manu-facturing method allows for more complex designs, especiallywhen compared to competing technologies such as high pre-cision micromachining or top-down wafer level manufacturing.The arbitrary and efficient redistribution of light is limited inthe case of a high spatial and angular extent of the source incomparison with the optics. Nonetheless, it is possible to createunconventional irradiance patterns as well. As an example,a design consisting of three split branches is created, leading

Fig. 4. Characterization. (a) Test setup using a Lambertian screen.(b) Cross section of measured and simulated irradiance. The measuredcurve of the bare chip is scaled together with the corresponding curve ofthe DTIRC which is normalized to its maximum. (c), (d) Comparisonof measured and simulated irradiance distributions, normalized to theirmaximum, respectively. (e) Normalized radiant intensity in polarrepresentation.

Letter Vol. 41, No. 13 / July 1 2016 / Optics Letters 3031

to an irradiance distribution with a triangular shape. Figure 5shows the comparison of simulated and measured irradiancepatterns, as well as the results of the 3D printing. The simu-lation and experiment show a good matching; the designedshape is well reproduced by the 3D printing.

In conclusion, we have successfully demonstrated a highlyminiaturized primary optics, directly manufactured on a LEDchip by femtosecond direct laser writing. To the best of ourknowledge, the presented device provides unprecedented colli-mation efficiency in terms of its size. The achieved resultsmatch well with theoretical considerations and exhibit high col-limation efficiency and high quality of the optics. The methodenables an arbitrary geometric complexity at a very small scaleand, thus, is perfectly suited for miniaturized devices. While thehigh writing times prevent a cost-effective mass manufacturingat the moment, small volume production becomes more attrac-tive due to the low initial cost. For the future, multipleapproaches such as shell writing or parallelization (using multi-ple beams through holographic beam splitters [22]) promise tohelp reducing the writing time drastically. We expect thismethod to be important for a number of applications suchas endoscopic lighting, proximity sensing, and communicationor signaling. Special use is assumed for rapid prototyping due tothe cheap and short design to manufacturing transfer.

Acknowledgment. The authors thank Dr. WolfgangHuber from “Chips 4 Light GmbH” for the generous supplyof LED chips. Special thanks go to Marion Hagel from MaxPlanck Institute for Solid State Reseach for rebonding the chips.The authors also acknowledge support from BW Stiftung(Spitzenforschung II), ERC (Complexplas), DFG, BMBF(PRINTOPTICS), and Carl Zeiss Stiftung.

REFERENCES

1. R. Blomaard and J. Biskop, NIP & Digital Fabrication Conference(2015).

2. A. Žukauskas, G. Batavičiūtė, M. Ščiuka, T. Jukna, A. Melninkaitis,and M. Malinauskas, Opt. Mater. Express 4, 1601 (2014).

3. M. Deubel, G. V. Freymann, M. Wegener, S. Pereira, K. Busch, andC. M. Soukoulis, Nat. Mater. 3, 444 (2004).

4. N. Lindenmann, G. Balthasar, D. Hillerkuss, R. Schmogrow, M.Jordan, J. Leuthold, W. Freude, and C. Koos, Opt. Express 20,17667 (2012).

5. Y. Li, Y. Yu, L. Guo, C. Chen, L. Niu, A. Li, and H. Yang, J. Opt. 12,035203 (2010).

6. T. Gissibl, S. Thiele, A. Herkommer, and H. Giessen, “Sub-micrometreaccurate free-form optics by three-dimensional printing on single-mode fibres,” Nat. Commun. (2016), doi: 10.1038/ncomms11763.

7. M. Malinauskas, A. Zukauskas, V. Purlys, K. Belazaras, A. Momot, D.Paipulas, R. Gadonas, A. Piskarskas, H. Gilbergs, A. Gaidukeviciute,I. Sakellari, M. Farsari, and S. Juodkazis, J. Opt. 12, 124010 (2010).

8. A. Žukauskas, M. Malinauskas, and E. Brasselet, Appl. Phys. Lett.103, 181122 (2013).

9. T. Gissibl, S. Thiele, A. Herkommer, and H. Giessen, “Two-photon di-rect laser writing of ultracompact multi-lens objectives,”Nat. Photonics(2016), doi: 10.1038/nphoton.2016.121.

10. S. Bianchi, V. P. Rajamanickam, L. Ferrara, E. Di Fabrizio, C.Liberale, and R. Di Leonardo, Opt. Lett. 38, 4935 (2013).

11. J. J. Wierer, M. R. Krames, J. E. Epler, N. F. Gardner, M. G. Craford, J.R. Wendt, J. A. Simmons, and M. M. Sigalas, Appl. Phys. Lett. 84,3885 (2004).

12. M. Ma, J. Cho, E. F. Schubert, Y. Park, G. B. Kim, and C. Sone, Appl.Phys. Lett. 101, 141105 (2012).

13. P. Schreiber, S. Kudaev, R. Steinkopf, and H. Schmidt,“Concentrators for high-power LEDs,” Fraunhofer Annual Report(2004), p. 29.

14. T. Jiang, Q. D. Chen, J. Zhang, Z. N. Tian, L. G. Niu, Q. S. Li, H. Y.Wang, L. Qin, and H. B. Sun, Opt. Lett. 38, 3739 (2013).

15. W. Mönch, Adv. Opt. Technol. 4, 79 (2015).16. J. Chaves, Introduction to Nonimaging Optics (CRC Press, 2008).17. W. T. Winston and R. Welford, The Optics of Non-imaging

Concentrators (Academic, 1978).18. http://www.nanoscribe.de/en/products/ip‑photoresists/.19. M. Ma, F. W. Mont, X. Yan, J. Cho, E. F. Schubert, G. B. Kim, and C.

Sone, Opt. Express 19, A1135 (2011).20. R. Wirth, C. Karnutsch, S. Kugler, and K. Streubel, IEEE Photon.

Technol. Lett. 13, 421 (2001).21. R. J. Koshel, Illumination Engineering (Wiley, 2013).22. L. Yang, A. El-Tamer, U. Hinze, J. Li, Y. Hu, W. Huang, J. Chu, and B.

Chichkov, Opt. Laser Eng. 70, 26 (2015).

Fig. 5. (a) Simulated irradiance distribution of a nonconventionalcollimator optics. (b) CAD representation. (c) Measured irradiancedistribution. (d) Microscope image of the structure. The black dottedlines in (a) and (c) are equal in size and are guides to the eye for bettercomparison.

3032 Vol. 41, No. 13 / July 1 2016 / Optics Letters Letter